Shapley value method improves feature attribution with missing multi-site data
Shapley Value Estimation for Multi-Site Data with Blockwise-Missing Features
Machine Learning
Summary
Assessing how much each feature contributes to a prediction is important, but often data from different places have different missing pieces. The authors show that simply filling in missing data can introduce errors in these assessments. They present a new method called FUSHAP that uses partially observed data from other sites without filling in missing values, reducing errors. Their method can also tell when data from other sites do not fit well with the target data, helping improve reliability.
What this means in practice
- •For hospital data teams: Improve accuracy of feature importance for clinical models using data from multiple hospitals with differing recorded features.
- •For environmental monitoring analysts: Enhance interpretation of air quality models by combining data from various monitoring sites that record different environmental variables.
Authors
Siqi Li, Wangxuan Fan, Yiming Li, Doudou Zhou, Molei Liu
Abstract
Shapley value (SV)-based methods are the prevailing framework for feature attribution in machine learning, yet existing population-level Shapley estimators generally assume that observations used to evaluate the coalitional game are fully observed under a common feature space. This assumption is routinely violated in multi-site studies across biomedicine, social science, and environmental monitoring, where institutions record different features under different protocols, producing systematic blockwise missingness across sources. We first show that the standard remedy of imputing missing features before computing Shapley values introduces systematic, coalition-dependent bias into the resulting attributions. We then propose \textbf{FUSHAP} (\textbf{Fu}sion \textbf{Sh}apley \textbf{A}ttribution from \textbf{P}artially-observed data), a method that leverages partially-observed auxiliary sites to reduce the variance of a preliminary single-site Shapley estimate without imputation. A permutation-based screening step detects and excludes sites whose data distributions are incompatible with the target population. In synthetic experiments, FUSHAP achieves $3$--$8\times$ lower MSE than the single-site estimator and $2$--$3\times$ lower MSE than imputation baselines without incurring imputation-induced bias, and the screening procedure identifies misaligned sites with $82\%$ power at moderate misalignment and $100\%$ for strong misalignment. On multi-site air quality and multi-center clinical data, FUSHAP reduces MSE by approximately $3$--$7\times$ relative to the single-site estimator; in the clinical application, standard imputation can increase MSE above the single-site baseline.